Academic journal article Applied Health Economics and Health Policy

Analytical Strategies for Characterizing Chemotherapy Diffusion with Patient-Level Population-Based Data

Academic journal article Applied Health Economics and Health Policy

Analytical Strategies for Characterizing Chemotherapy Diffusion with Patient-Level Population-Based Data

Article excerpt

Understanding factors that drive technology diffusion facilitates systematic planning by healthcare delivery systems. Identification of geographical regions or patient subsets in which uptake is delayed may reveal barriers to optimal use and prioritize areas for future clinical trials. Estimation of diffusion can also be helpful when a new drug is suspected as a cause of an uncommon but potentially serious late effect.[1,2]

In cancer, new therapeutic agents earn US FDA approval based on measures of safety and efficacy from clinical trials that involve only limited numbers of patients. FDA approval typically includes some specific criteria dictating circumstances in which use of a new drug has been found to be beneficial. However, the attributes of patients in the general cancer population are not representative of the research subjects, which leads to great variation in how new cancer treatments are used. There is evidence of both underuse (failure to administer treatment in circumstances where it has proven efficacy)[3] and overuse (administration of new treatments in circumstances where efficacy has not yet been established),[4] with factors related to patients, their cancers, their physicians and of course the treatments themselves influencing the diffusion process. Because it is of particular interest to evaluate how new treatments, and especially new chemotherapy agents, are incorporated into practice,[5] we describe approaches for characterizing this process. We demonstrate how existing statistical methods can be extended or adapted to the study of chemotherapy diffusion.

While pre-existing research investigates patterns of non-cancer drug dissemination at the physician level,[6,7] the goal of this article is to set a framework for characterizing diffusion of new chemotherapeutic agents at the patient level. To place our research in the context of the relevant economics and healthcare literature, we review existing empirical methods and evaluate their appropriateness for investigating chemotherapy diffusion. In the setting of advanced cancer, or any other condition with short survival, where people are eligible to adopt an innovation for only a short time interval, existing methods do not easily capture the short-term dynamics of diffusion. In addition, patient-level population-based data pose several analytical challenges.

First, information on cancer diagnosis is collected on a continuous basis, with the addition of newly diagnosed patients regularly enriching the pool of potential drug users. This generates staggered entry of patients into the dataset. Second, the complicated relationship between the probability of receiving the drug and the patient's risk of death generates a competing risk problem. Third, when attempting to understand the forces underlying chemotherapy diffusion, there are two distinct populations of adopters that can serve as potential units of analysis. The first is the population of oncologists who use their expertise to decide whether a given drug is medically appropriate in a particular circumstance. The second is the population of cancer patients who stand to benefit from adoption. When the unit of analysis is the cancer patient, the analysis must account for the effect of clustering within physician that arises from the fact that practice patterns within particular clinics or physicians are likely to be correlated. Our proposed method is equipped to address all these issues. We illustrate this approach with two case studies relying on linked Surveillance, Epidemiology, and End Results (SEER)-Medicare data to characterize the uptake of gemcitabine, FDA approved for pancreatic cancer in 1996, and of irinotecan, FDA approved for second-line treatment of metastatic colorectal cancer, also in 1996. We describe analytical approaches for characterizing drug diffusion when patient-level population-based data are available. In these analyses, we consider the cancer patient as the unit of analysis, while accounting for the fact that practice patterns within particular clinics or physicians are correlated. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.